Poly(A) Tail Length Is Controlled by the NuclearPoly(A)-binding Protein Regulating the Interactionbetween Poly(A) Polymerase and the Cleavage andPolyadenylation Specificity Factor*□S

Received for publication, May 7, 2009, and in revised form, June 4, 2009 Published, JBC Papers in Press, June 9, 2009, DOI 10.1074/jbc.M109.018226

Uwe Kuhn, Miriam Gundel1, Anne Knoth2, Yvonne Kerwitz3, Sabine Rudel4, and Elmar Wahle5

From the Institute of Biochemistry and Biotechnology, Martin-Luther-University Halle-Wittenberg, Kurt-Mothes-Strasse 3,06120 Halle, Germany

Poly(A) tails of mRNAs are synthesized in the cell nucleuswith a defined length, �250 nucleotides in mammalian cells.The same type of length control is seen in an in vitro polyad-enylation system reconstituted from three proteins: poly(A) po-lymerase, cleavage and polyadenylation specificity factor(CPSF), and the nuclear poly(A)-binding protein (PABPN1).CPSF, binding the polyadenylation signal AAUAAA, andPABPN1, binding the growing poly(A) tail, cooperatively stim-ulate poly(A) polymerase such that a complete poly(A) tail issynthesized in one processive event, which terminates at alength of�250 nucleotides.We report that PABPN1 is requiredto restrict CPSF binding to the AAUAAA sequence and to per-mit the stimulation of poly(A) polymerase by AAUAAA-boundCPSF to bemaintained throughout the elongation reaction. Thestimulation by CPSF is disrupted when the poly(A) tail hasreached a length of �250 nucleotides, and this terminates pro-cessive elongation. PABPN1measures the length of the tail andis responsible for disrupting the CPSF-poly(A) polymeraseinteraction.

Thepoly(A) tails present at the 3� end of almost all eukaryoticmRNAs have two major functions. The first function is in thecontrol of mRNA decay; degradation of the poly(A) tail by a 3�exonuclease (deadenylation) is the first step in both of the twomain pathways of mRNA decay, and the completion of dead-enylation triggers the second step, either cap hydrolysis or fur-ther 3�–5� degradation. Because the rate of deadenylation isgoverned by sequence elements in the mRNA, it is specific foreach mRNA species and serves as a major determinant of

mRNA half-life (1–3). Obviously, a control of mRNA stabilityby the rate of deadenylation requires a defined poly(A) length asa starting point. The second function of the poly(A) tail is inthe initiation of translation; the cytoplasmic poly(A)-bind-ing protein associated with the poly(A) tail promotes theinitiation of translation by an interaction with the initiationfactor eIF4G and probably through additional mechanisms(4–7). In this process, poly(A) tail length can also be impor-tant. For example, gene regulation during oocyte maturationand early embryonic development of animals depends ontranslational regulation of maternal mRNAs, and changes inpoly(A) tail lengths of specific mRNAs, determined both bydeadenylation and by regulated cytoplasmic poly(A) exten-sion, play a major role in this translational regulation. Longpoly(A) tails favor translation, whereas a shortening of thetail promotes translational inactivation of the message (8, 9).Similar mechanisms seem to operate in neurons (10, 11) andpossibly in other somatic cells (12).Because the length of the poly(A) tail is important for its

function, it is not surprising that poly(A) tails are generally syn-thesized with a defined length, which is species-specific,�70–90 nucleotides in Saccharomyces cerevisiae (13, 14) and�250 nucleotides in mammalian cells (15). Subtle differencesbetween newly made poly(A) tails of different mRNAs havebeen described (13), and there is even a class of mRNAs thatnever receives more than an oligo(A) tail (16, 17). However, theheterogeneous length distribution seen in the steady-statemRNApopulation is the result of cytoplasmic shortening start-ing from a relatively well defined initial tail length; heterogene-ity of tail length reflects age differences of themRNAmolecules.The oligo(A) tails present on inactive mRNAs in oocytes orembryos are also generated by shortening of full-length tailsmade in the cell nucleus (18).The poly(A) tail is added during 3� end processing of mRNA

precursors in the cell nucleus (19–21). This reaction consists oftwo steps: an endonucleolytic cleavage followed by the additionof the poly(A) tail to the upstream cleavage product.Whereas alarge proteinmachinery of some 20 ormore polypeptides (22) isrequired for the cleavage reaction, subsequent polyadenylationhasmuch simpler protein requirements. In themammalian sys-tem, it can be reconstituted from three proteins: poly(A) poly-merase, the enzyme catalyzing primer-dependent polymeriza-tion ofAMPusingATP as a precursor (23–25); the cleavage and

* This work was supported by a grant from the Deutsche Forschungsgemeinschaft.□S The on-line version of this article (available at http://www.jbc.org) contains

supplemental Table S1 and Figs. S1 and S2.1 Present address: Max Planck Research Unit for Enzymology of Protein Fold-

ing, Weinbergweg 22, 06120 Halle, Germany.2 Present address: Institute of Biochemistry and Molecular Biology, University

of Freiburg, Stefan-Meier-Strasse 17, 79104 Freiburg, Germany.3 Present address: MicroCoat Biotechnologie GmbH, Am Neuland 3, 82347

Bernried, Germany.4 Present address: Max Planck Institute of Biochemistry, 82152 Martinsried,

Germany.5 To whom correspondence should be addressed: Kurt-Mothes-Str. 3, 06120

Halle, Germany. Tel.: 49-345-5524920; Fax: 49-345-5524920; E-mail:ewahle@biochemtech.uni-halle.de.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 284, NO. 34, pp. 22803–22814, August 21, 2009© 2009 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

AUGUST 21, 2009 • VOLUME 284 • NUMBER 34 JOURNAL OF BIOLOGICAL CHEMISTRY 22803

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

polyadenylation specificity factor (CPSF),6 which binds thecleavage and polyadenylation signal AAUAAA (26, 27); andthe nuclear poly(A)-binding protein (PABPN1), which bindsthe growing poly(A) tail (28, 29). Note that PABPN1 is distinctfrom the family of cytoplasmic poly(A)-binding proteins (30).Roles of poly(A) polymerase and CPSF in polyadenylation invivo have beenmost clearly demonstrated by genetic analysis ofthe orthologues in S. cerevisiae (21, 31). PABPN1 has no func-tional orthologue in budding yeast (32); its function in polyad-enylation has been confirmed in mammalian cells (33) and inDrosophila (34).Whereas PABPN1 and poly(A) polymerase are monomeric

proteins, CPSF is a hetero-oligomer, which has not yet beenreconstituted from recombinant proteins (22, 26, 35–40).Poly(A) polymerase on its own is barely active because of a lowaffinity for its RNA substrate and thus acts distributively, i.e. itdissociates from the RNA after each polymerization step, andpresumably often before it has incorporated any nucleotide; theenzyme also has no significant sequence specificity and willelongate any RNA with a free 3� OH (24). Both CPSF andPABPN1 enhance the activity of the polymerase by recruitingthe enzyme to its substrate through direct interactions (38, 41).Sequence specificity of poly(A) addition reflects the RNA bind-ing specificities of the two stimulatory factors: CPSF recruitsthe polymerase to RNAs containing the AAUAAA sequencein the vicinity of their 3� ends (24, 42, 43), and PABPN1 recruitsthe enzyme to substrate RNAs carrying a terminal oligo(A)tract (29). Each factor alone endows the polymerase with mod-est processivity, such that it can incorporate maybe two to fivenucleotides before dissociating (44). RNAs containing both theAAUAAA sequence and an oligo(A) tail and thus resemblingintermediates of the polyadenylation reaction support a coop-erative or synergistic stimulation of poly(A) polymerase by bothCPSF and PABPN1. Under these conditions, addition of thepoly(A) tail occurs in a processive manner, i.e. without inter-mittent dissociation of the protein complex from its substrateRNA (29, 44).Interestingly, the reconstituted polyadenylation reaction

also shows proper length control, generating poly(A) tails of thesame length as seen in vivo; tails grow to a relativelywell definedlength of 250–300 nucleotides in a rapid, processive reaction(29, 44). Length control is due to termination of this processiveelongation; extension beyond 250 A residues is largely distrib-utive and therefore slow (45). These kinetics of in vitro poly(A)tail synthesis are fully consistent with the in vivo kineticsderived from pulse-labeling studies (46). In vitro, poly(A) tailelongation rates beyond 250 A residues are similar when eitherCPSF or PABPN1 or both are present. In other words, sub-strateswith long poly(A) tails no longer support the cooperativestimulation of poly(A) polymerase by both CPSF and PABPN1that is the basis of processive elongation (45). The terminationof processive elongation must be mediated by a change in theRNA-protein complex that remains to be defined.When RNAscarrying poly(A) tails of different lengths are used as substrates

for polyadenylation, the tails are always elongated processivelyto 250 nucleotides, independently of the initial length, whereasextension of a tail of 250 or more nucleotides in length is slowanddistributive from the start of the reaction. Thus, poly(A) taillength control is based on some kind of AMP residue countingor length measurement, not on a kinetic mechanism (45).In this paper, we address the two problems outlined above:

first, how does the polyadenylation complex change to termi-nate processive poly(A) tail elongation, and second, how is thelength of the tail measured?We provide evidence that PABPN1is the active component in the mechanism of length control.The protein promotes the interaction between CPSF andpoly(A) polymerase when bound to a short poly(A) tail.PABPN1 no longer promotes or even actively disrupts thisinteraction when bound to a poly(A) tail of 250 nucleotides orlonger and thereby terminates the cooperative, processive elon-gation reaction in a poly(A) tail length-dependent manner.Only poly(A) sequences are counted as part of the tail. Becausethis reflects the binding specificity of PABPN1 and because dis-ruption of the CPSF-poly(A) polymerase interaction requirescomplete coverage of the poly(A) tail by this protein, PABPN1is also the protein that measures the length of the tail.

EXPERIMENTAL PROCEDURES

Plasmids—The expression plasmid containing a partiallysynthetic PABPN1 gene has been described (47). The LALAdouble mutant was generated by introduction of the L119Amutation into the plasmid pGM-synPABPN1 L136A (41). ThePABPN1 open reading frame was then subcloned with NdeIandBamHI into the expression vector pUK (47). For expressionof poly(A) polymerase, the PAP open reading frame frompGM10-PAP82 (48) was subcloned into the pUK plasmid withNdeI and BamHI.Plasmids encoding various polyadenylated derivatives of the

L3pre RNA have been described (41). In addition, pSP64-L3preA75 was generated by the same procedure. The mutagen-esis of the polyadenylation signal (� variant: AAGAAA) ofpSP64-L3preA15 and derivatives (41) was done by PCR using �forward and reverse primers (supplemental Table S1) and PwoDNA polymerase (Peqlab). The PCR fragment was digestedwith PstI and PvuII and subcloned into the PstI/PvuII openedplasmids.RNA—A280 was obtained by size fractionation of commer-

cial poly(A) and labeled as described (49). Poly(A) concen-trations were determined by UV spectroscopy with �(AMP) � 9800 cm�1 M�1.

Other RNAs were synthesized by run-off transcription withSP6 RNApolymerase (RocheApplied Science) and [�-32P]UTP(GE Healthcare or Hartmann Analytics). No 5� cap wasadded. RNAs were gel-purified and quantified as described(45). Templates for run-off transcription were either plas-mid DNA-digested with appropriate restriction enzymes orPCR fragments that were amplified with an SP6 promotorforward primer (supplemental Table S1) and a reverseprimer that hybridized downstream to the plasmid. DNAtemplates were purified by phenol/chloroform extractionfollowed by ethanol precipitation.

6 The abbreviations used are: CPSF, cleavage and polyadenylation specificityfactor; PABPN1, poly(A)-binding protein nuclear 1; nt, nucleotide(s); PAP,poly(A) polymerase.

Poly(A) Tail Length Control

22804 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 34 • AUGUST 21, 2009

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

Polyadenylated RNAs L3preA15, L3preA45, L3preA75, andL3preA105 and the corresponding versions with a point muta-tion in AAUAAA were generated by transcription of suitableplasmids templates (41) as above. RNAs with poly(A) tails lon-ger than 105 nucleotides were obtained by further elongation ofL3preA15 or similar RNAs as described (45). The RNA sub-strates L3pre-N49A15 and L3preA15N63A15 were synthesizedfrom PCR fragments generated as described above using theplasmids pSP64-L3pre and pSP64-L3preA15 as templates andthe SP6 promotor primer and the corresponding reverse prim-ers (supplemental Table S1).A first set of RNA processing substrates with no poly(A) tail

and with increasing distance between the AAUAAA sequenceand the 3� end were generated by transcription from the plas-mids pSP64-L3 or pSP64-L3pre� linearized with RsaI close tothe 3� processing cleavage site (1 nt upstream of the cleavagesite, 19 nt downstream of AAUAAA) or further downstreamwith DraI (48 nt downstream of the cleavage site), AflIII (101nt), EcoRI (162 nt), andMbiI (205 nt) and from a PCR fragment(299 nt distance) generated by use of the SP6 promotor primerand a second oligonucleotide priming about 300 nt down-stream of the RsaI site (L299 rev; supplemental Table S1). Asecond set of processing substrates with increasing distancebetween the AAUAAA sequence and the 3� end was synthe-sized from PCR fragments generated from the templatespSP64-L3 and pSP64-L3pre�, respectively, with the help of theSP6 promotor primer and different downstream primers (sup-plemental Table S1). The downstream primers were designedsuch that all RNAs carried the same additional sequence(-UGUA) at the 3� end.When RNAdraw (54) predicted that the3� end of a particular RNA was included in a double-strandedregion, the additional 3� terminal sequence was extended toUGUUGUA to disrupt the double-stranded structure.Proteins—Expression of His-tagged PABPN1 LALA was

done as described (47) with the following modifications: M9minimalmediumwas used, and inductionwas started atA600 �1 by addition of 0.4 mM isopropyl �-D-thiogalactopyranoside at25–30 °C for 3 h. The cells were harvested by centrifugation,resuspended in 50mMsodiumphosphate (pH8.0), 300mMKCl,10% glycerol, and lysed with a Basic Z cell disruptor (ConstantSystems Ltd.) at 2200 bar. Nickel-nitrilotriacetic acid chroma-tographywas performed as described (47). Eluted proteinsweredialyzed first against RS buffer (50 mM sodium phosphate, pH6.5, 10% glycerol, 1 mM dithiothreitol, 1 mM EDTA) containing1 M NaCl and then against RS buffer containing 300 mM salt.After dilution to a salt concentration below 150 mM, nucleicacids and contaminating proteinswere removed by chromatog-raphy on a 1-ml Resource S column (GEHealthcare).Wild-typePABPN1was expressed in a fermentor-grown culture and puri-fied by a similar procedure. Because recombinant His-taggedPABPN1 is not always fully active, the amount required to sat-urate the poly(A) tails of substrate RNAs was determined bytitrations in gelmobility shift assays. For some experiments, calfthymus PABPN1 was used (50).CPSF was purified essentially as described (45). Full-length

bovine poly(A) polymerase was also expressed in Escherichiacoli Bl21 pUBS. Bacteria were grown in SOBmedium toA600 �1 and induced by 1 mM isopropyl �-D-thiogalactopyranoside.

The cells were harvested after 16 h at 25–30 °C and disrupted asabove. TheHis-tagged PAPwas purified by nickel-nitrilotriace-tic acid chromatography (Qiagen) followed by a Hitrap Hepa-rin-Sepharose and a MonoQ column (GE Healthcare).Concentrations of PABPN1 and poly(A) polymerase were

determined from the absorption at 280 nm, and extinctioncoefficients were calculated by the Lasergene software package(DNAStar Inc.). The concentration of CPSF was estimated bytitrations in gel mobility shift and polyadenylation assays,assuming that a 1:1 ratio to the substrate RNA was required tosaturate the reaction.Polyadenylation Assays—Polyadenylation assays were per-

formed at 37 °C as described (45) except that 2% (w/v) polyeth-ylene glycol 6000 was substituted for polyvinyl alcohol in thereaction buffer. The reaction volume was 20 or 25 �l or multi-ples thereofwhen several time pointswere taken fromone reac-tion. RNA substrates, proteins, and incubation times are indi-cated in the figure legends. Amounts of RNA and proteinalways refer to the unit reaction volume of 20 or 25 �l. Molarquantities of RNA refer to polymers, notmononucleotides. Thereaction products were resolved on 5 or 6% polyacrylamide gelscontaining 8.3 M urea, and radioactivity was detected by phos-phorimaging. Polyadenylation assays often included a reactioncontaining a 10-fold amount of poly(A) polymerase to controlfor the ability of the RNA to be elongated and to gauge thestimulatory effects of the RNA binding proteins.Quantitation of PAP Stimulation by CPSF—Polyadenylation

reactions were performed with 80 fmol of radioactively labeledRNA, 80 fmol of PAP, and an appropriate amount of CPSF(approximately equimolar to the RNA) in the presence of 1.25�g of competitor tRNA for 0, 2, and 10 min at 37 °C. To meas-ure the activity of unstimulated PAP, control reactionswith 800fmol of PAP were performed in the absence of CPSF for 10 minat 37 °C. Polyadenylation products were analyzed as above andquantified with ImageQuant 5.0 (GE Healthcare). Efficiency ofpolyadenylation was determined as follows: for each lane, a linegraph was integrated in Excel (Microsoft Corp.) to determinethe total amount of radioactivity. Then the position in the gelwas determined that divided the lane into two portions eachcontaining the same amount of radioactivity. DNA size stand-ards in neighboring lanes were used to determine the RNAlength corresponding to this position. The elongation of anRNA is given by the difference between this position and that ofthe sameRNAwithout incubation (t� 0min). PAP stimulationwas calculated as the ratio of stimulated elongation (in the pres-ence of CPSF) over unstimulated elongation (in the absence ofCPSF). Because extension by 1� poly(A) polymerase (80 fmol)was barely detectable, stimulation was normalized to 1⁄10 of theextension seen in the 10�poly(A) polymerase lane. Because theRNA intensities are not weighted by the length, this type ofquantitationmeasures mostly the quantity of RNA that is elon-gated and puts less emphasis on the length of poly(A) that isadded. This is desired because the former parameter moreaccurately reflects the AAUAAA-dependent stimulation byCPSF.Electrophoretic Mobility Shift Assays—80 fmol of RNA was

incubated under reaction conditions in the absence of ATP.The reaction volume was 20 �l, and proteins were used as indi-

Poly(A) Tail Length Control

AUGUST 21, 2009 • VOLUME 284 • NUMBER 34 JOURNAL OF BIOLOGICAL CHEMISTRY 22805

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

cated in the figure legend. After 30 min of incubation at roomtemperature, protein-RNA complexes were resolved by nativegel electrophoresis as described (49).

RESULTS

PABPN1 Maintains the AAUAAA Dependence of CPSFBinding—With RNA substrates carrying no poly(A) tails, thestimulation of polyadenylation by CPSF is strongly dependenton the presence of the CPSF-binding site, AAUAAA. Becauseof the similarity of this sequence with poly(A) and becauseCPSF binds tightly to a poly(A) column (26), the possibility wastested that CPSF could stimulate polyadenylation also by bind-ing to the growing poly(A) tail. Indeed, CPSF stimulated exten-sion of size-fractionated poly(A) by poly(A) polymerase quiteefficiently, although less so than PABPN1 (Fig. 1A; activity ofthe same proteins on the regular polyadenylation substrate isshown as a control in the first five lanes). When PABPN1 wastitrated into these poly(A) elongation reactions, the rate ofelongation increased, but the CPSF-dependent stimulation wasreduced to zero at a concentration of PABPN1 sufficient tocover the poly(A) substrate completely (Fig. 1B, compare lanes

plus and minus CPSF). Thus, CPSFcan stimulate poly(A) polymerasewhen bound to naked poly(A), but itcannot bind the poly(A)-PABPN1complex. The effect of CPSF wasthen examined with a “regular” pol-yadenylation substrate carrying anAAUAAA sequence; the L3preRNA corresponds to the cleavedintermediate of a 3�processing reac-tion at the L3 site of the adenovirusmajor late transcript. In L3pre�, thepolyadenylation signal ismutated toAAGAAA. Variants of these RNAswere used that carried differentlengths of 3� terminal poly(A)already at the start of the elongationreaction, thus representing inter-mediates in the elongation reaction.In the absence of PABPN1, CPSFstimulation of the extension of suchsubstrates became increasinglyinsensitive to a point mutation inthe AAUAAA sequence (L3pre�substrate) with increasing poly(A)tail length; AAUAAA dependencewas pronounced with an A15 tail(Fig. 2, compare lanes 7–9 withlanes 10–12), weak with an A105 tail(lanes 19–21 versus lanes 22–24),and absent with an A190 tail (lanes31–33 versus lanes 34–36). Thisconfirms stimulation of poly(A) po-lymerase by CPSF bound to poly(A).The addition of PABPN1 restoredAAUAAAdependence (Fig. 2; com-pare, for example, lanes 25–30 with

lanes 31–36). Therefore, in all subsequent experiments, a com-plete coverage of the poly(A) tail with PABPN1was ascertainedby titration experiments analyzed by gel mobility shifts. Thesecontrols guaranteed that any stimulation byCPSFwas due to itsoccupancy of the AAUAAA site.PABPN1 Facilitates the Stimulation of Poly(A) Polymerase by

CPSF—We have previously reported that CPSF can stimulatepoly(A) polymerase efficiently even on RNAs carrying longpoly(A) tails (45). The results described in the preceding sectionnow suggest that thismay have been due toCPSF binding to thepoly(A) tail. Therefore, RNAswithout a poly(A) tail andwith anincreasing distance between the AAUAAA sequence and the 3�end were tested for their ability to support the stimulation ofpoly(A) polymerase by CPSF. Two such series of RNA sub-strates were used. Because poly(A) polymerase is sensitive tothe type of nucleotides at the 3� end, the second series of RNAswas designed to have a uniform3� end sequence (-UGUA). Twocorresponding series of RNAs carrying a point mutation in theAAUAAA sequence served as controls. With the two mutantseries of substrates, CPSF stimulation was weak and essentiallyindependent of the length of the RNA (Fig. 3; an example of raw

FIGURE 1. Stimulation of poly(A) polymerase by CPSF bound to straight poly(A) is sensitive to PABPN1.A, CPSF can stimulate poly(A) polymerase when bound to straight poly(A). 80 fmol of L3preA15 or A100 wasincubated in a 25-�l reaction with an approximately equimolar amount of CPSF, 20 fmol of PAP and 1.6 pmolof His-PABPN1 as indicated. After 2 min of preincubation at 37 °C, the reactions were started by the addition ofATP. The reactions were stopped after 30 s or 15 min as indicated, and RNAs were analyzed by gel electro-phoresis. The size of DNA markers in nucleotides is indicated on the left. B, PABPN1 inhibits the stimulatoryactivity of CPSF on straight poly(A). 80 fmol of A280 was incubated in 20-�l reactions with an approximatelyequimolar amount of CPSF, 32 fmol of PAP, and increasing amounts of PABPN1 as indicated. 1.5 pmol ofPABPN1 was sufficient for A280 coverage as judged by an independent gel mobility shift assay. The reactionswere preincubated for 10 or 15 min at 37 °C and started by addition of ATP. They were stopped after 15 min andanalyzed by gel electrophoresis. A control reaction containing a 10-fold amount of poly(A) polymerase wasincluded as explained under “Experimental Procedures.”

Poly(A) Tail Length Control

22806 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 34 • AUGUST 21, 2009

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

data is shown in supplemental Fig. S1). With the wild-typeseries of substrates, the stimulatory effect of CPSF was strongwhen the RNA ended at the normal cleavage site and gotweaker with an increasing distance. The decrease in polyad-enylation efficiencywasmore gradual in one series of RNAs andmore precipitous in the other (Fig. 3 and supplemental Fig. S1).In contrast to this pronounced distance sensitivity of the CPSFeffect, the rate of the processive elongation reaction in the pres-ence of both CPSF and PABPN1 remains constant at �25 nt/sall the way up to a tail length of 250 nucleotides (Fig. 3 in Ref.45). From the efficiency of elongation we infer that, in the reac-tion lacking PABPN1, the interaction between CPSF andpoly(A) polymerase is progressively weakened with an increas-

ing distance of the two proteins onthe RNA, whereas in the presence ofPABPN1, the interaction is main-tained over a distance of up to 250nucleotides. In otherwords, PABPN1does not merely passively permitCPSF to interact with poly(A) poly-merase; rather, PABPN1 actively pro-motes this interaction, presumably byfacilitating a folding back of the RNA.With increasing distance between the3� endundergoing elongation and theCPSF-binding site, the CPSF-poly(A)polymerase interaction becomesincreasingly dependent on PABPN1.Both CPSF and PABPN1 Remain

Associated with the RNA Substrateafter Termination of ProcessivePolyadenylation—How does thepolyadenylation complex change toterminate processive elongation?Cooperative stimulation of poly(A)polymerase byCPSF andPABPN1 isessential for processive elongation,and the elongation beyond �250nucleotides proceeds at a rate char-acteristic for the stimulation byeither factor alone (45). Thus, ter-mination of the processive reactionmight be caused by the dissociationor displacement of one of the twostimulatory proteins. For example,PABPN1 might displace CPSF fromthe AAUAAA sequence because ofthe similarity of the binding site.This hypothesis was tested by gelshift experiments. PABPN1 wasused at a concentration sufficient tosaturate the L3preA300 substrate(Fig. 4, lane 3). In the absence ofPABPN1, 80% of the RNA werebound by CPSF (Fig. 4, lanes 4 and5). When PABPN1 was added inaddition, �50% of the PABPN1-as-sociated RNA bound CPSF simulta-

neously, as shown by the supershift (Fig. 4, lanes 7 and 8). Con-trols with the L3pre�A300 substrate showed that binding ofCPSF was AAUAAA-independent (Fig. 4, compare lanes 12and 13 with lanes 4 and 5), confirming binding to the poly(A)tail as seen in the functional assays reported above (Figs. 1 and2). Binding of CPSF to the poly(A) tail was prevented by itssaturation with PABPN1 (Fig. 4, compare lanes 9 and 10 withlanes 7 and 8). The experiment shows that RNAs with a longpoly(A) tail that can no longer undergo processive elongationcan nevertheless bind CPSF and PABPN1 simultaneously.The binding of CPSF to RNAs carrying different poly(A) tail

lengths was also examined by UV cross-linking experiments.Although cross-linking of the 30-kDa subunit of CPSF was very

FIGURE 2. PABPN1 maintains the AAUAA dependence of CPSF stimulation. The reactions were set upcontaining 80 fmol of RNA/25 �l of unit volume, either wild type (wt) or with a point mutation in the polyad-enylation signal (�) and with different poly(A) tail lengths, as indicated. The amounts of CPSF were approxi-mately equimolar to the RNA, 20 fmol of poly(A) polymerase, and 1.6 pmol PABPN1 was used as indicated. Thereaction mixtures were prewarmed for 3 min at 37 °C, the samples were withdrawn for the 0 min time points,and the reactions were started by the addition of ATP. Aliquots were taken at the time points indicated, andRNA was analyzed by gel electrophoresis. The sizes of DNA markers are indicated on both sides.

Poly(A) Tail Length Control

AUGUST 21, 2009 • VOLUME 284 • NUMBER 34 JOURNAL OF BIOLOGICAL CHEMISTRY 22807

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

inefficient under all conditions tested, there was only a slightdecrease when RNAs with long poly(A) tails and saturatingconcentrations of PABPN1 were used (supplemental Fig. S2).From both the gel shift and the cross-linking experiments, weconclude that a dissociation or displacement of CPSF does notplay a role in the termination of processive polyadenylation.PABPN1 Disrupts the Stimulation of Poly(A) Polymerase by

CPSF to Terminate Processive Polyadenylation—An alternativemechanism to terminate processive elongation might involve adisruption of the interaction between poly(A) polymerase andone of the two stimulatory factors without a displacement ofeither protein. It seems plausible that it is the interaction withCPSF that is lost; the CPSF stimulation gets progressivelyweaker with increasing distance between theCPSF-binding siteand the RNA 3� end (see above), presumably because the inter-action between CPSF and the enzyme can bemaintained only ifthe RNA folds back on itself. In contrast, the stimulation byPABPN1 is not affected by increasing poly(A) tail length (45);the activity of poly(A) polymerase continuously creates newbinding sites for PABPN1, and PABPN1 must remain close tothe polymerase (41). Thus, an attractive model for length con-trol would be that there is a length-dependent conformationaltransition in the poly(A)-PABPN1 complex such that PABPN1no longer facilitates or even actively disrupts the contactbetween CPSF and the polymerase. According to this model,the relatively slow elongation taking place after termination ofprocessive elongation should then be due to the remaining

stimulation of poly(A) polymerase by just PABPN1. That therate of extension of long poly(A) tails in the presence ofPABPN1 is insensitive to the presence or absence of CPSF (45)is consistent with this model but does not prove it. However,the model makes a simple but very specific prediction; if, onlong tails, PABPN1 stimulates poly(A) polymerase in the nor-mal fashion but disrupts the stimulation by CPSF, then amuta-tion in PABPN1 making this protein unable to stimulatepoly(A) polymerase should uncover its inhibitory influence onthe CPSF-poly(A) polymerase interaction. Moreover, thisinhibitory activity should be visible only on long tails, whereasthe mutant PABPN1 should still support the CPSF-poly(A) po-lymerase interaction on short tails.Point mutations in the hydrophobic face of the coiled-coil

domain of PABPN1,most strongly L136S or L136A, abolish theability of the protein to stimulate poly(A) polymerase withoutaffecting poly(A) binding (41). The corresponding mutation ofthe Drosophila orthologue of PABPN1, I61S, fails to rescue anull allele (34). For the purpose of testing the proposed model,we used the double mutant L119A/L136A (LALA for short). Incontrol reactions with a substrate RNA carrying a short oli-go(A) tail (L3preA15), the mutant protein showed a barelydetectable stimulation of poly(A) polymerase in the absence ofCPSF over a time course of 5 min (Fig. 5, compare lanes 17–20with lanes 1–4). In the presence of both CPSF and PABPN1LALA, the rate of elongation was very similar to that seen with

FIGURE 3. CPSF stimulation is sensitive to the distance between AAUAAAand the 3� end. The stimulation of poly(A) polymerase by CPSF was meas-ured with two sets of RNA substrates with an increasing distance between theCPSF-binding site AAUAAA and the 3� end. In each case, an identical set ofRNAs with a point mutation in AAUAAA was used as a control. Polyadenyl-ation reactions contained, per unit volume of 25 �l, 80 fmol of RNA, and 80fmol of PAP in the presence or absence of CPSF (amount approximatelyequimolar with RNA). After warming to 37 °C, aliquots for the 0-min timepoint were withdrawn, the reactions were started by the addition of ATP, andadditional aliquots were withdrawn after 2 and 10 min. A control reactioncontaining a 10-fold higher amount of poly(A) polymerase (800 fmol) wasincubated for 10 min in the absence of CPSF. PAP stimulation was analyzed asdescribed under “Experimental Procedures” based on the 10-min time points.The plotted data were averaged from at least two experiments for each set ofRNAs. Raw data for the first series of RNAs and the mutant controls are shownin supplemental Fig. S1. A, distance-dependent stimulation with the first setof RNA substrates having different 3� end sequences. The black squares rep-resent the data for RNAs with a wild-type AAUAAA sequence and show adecreasing elongation efficiency in proportion to an increasing distancebetween AAUAAA sequence and 3� end. Open squares represent the controldata obtained with the corresponding RNAs having a point mutation in theAAUAAA sequence. Elongation is less efficient than with the wild-typesequence and essentially independent of the length of the RNA. B, distance-dependent stimulation with the second set of RNA substrates having uniform3� end sequences (-UGUA). The data for the wild-type and mutant RNAs areshown as in A.

FIGURE 4. CPSF and PABPN1 can bind simultaneously to an RNA with along poly(A) tail. 80 fmol of L3pre-A300 orL3pre�-A300 was incubated in a20-�l reaction volume in the absence of ATP with 100 fmol of PAP, 2400 fmolof calf thymus PABPN1 (saturating as judged by titration in an independentgel mobility shift assay) and CPSF (approximately equimolar with RNA) in thepresence of 1.5 �g of competitor tRNA. After 30 min of incubation at roomtemperature, complexes were resolved on a native polyacrylamide/agarosecomposite gel. The reactions containing all three proteins were carried outtwice (lanes 7 and 8 and lanes 12 and 13, respectively).

Poly(A) Tail Length Control

22808 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 34 • AUGUST 21, 2009

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

CPSF alone; in other words themutant PABPN1 behaved as if itwere absent (Fig. 5, compare lanes 21–24 with lanes 5–8) (41).In contrast, a second control reaction with wild-type PABPN1and CPSF showed the expected processive elongation (Fig. 5,compare lanes 13–16 with lanes 5–8 and 9–12). With an RNAsubstrate carrying an A330 tail, 60-min time courses were usedbecause of the low elongation rates caused by the lack of pro-cessivity. Both wild-type PABPN1 and CPSF stimulated exten-sion individually, as expected (Fig. 5, compare lanes 33–36 and29–32 with lanes 25–28). Elongation in the presence of both

was very similar to the reaction seenwith PABPN1 alone, demonstratingthe absence of a cooperative effect,also as expected (Fig. 5, lanes37–40). PABPN1 LALA by itselfwas again unable to stimulatepoly(A) extension to any significantextent (Fig. 5, lanes 41–44).Remarkably, however, when themutant PABPN1 was used togetherwith CPSF, no stimulation of poly-adenylation was seen, i.e. the stimu-lation by CPSF alone was sup-pressed (Fig. 5, lanes 45–48), andthe rate of elongation was very sim-ilar to that caused by poly(A) poly-merase alone (lanes 25–28). Theseresults provide very strong supportfor the model outlined above. On along poly(A) tail, PABPN1 disruptsthe interaction between CPSF andpoly(A) polymerase.With wild-typeprotein, this is not evident becausePABPN1 itself stimulates poly(A)tail extension. The inhibitory effectof PABPN1 on the CPSF-poly(A)polymerase interaction becomesvisible when mutant PABPN1 isused that cannot stimulate polyad-enylation itself. Disruption of theCPSF-poly(A) polymerase interac-tion is limited to long poly(A) tails;the interaction is normal on shorttails.Because thismodel is proposed to

explain length control, the depend-ence of the switch in PABPN1 activ-ity on the length of the poly(A) tail isa central feature. Therefore, addi-tional poly(A) tail lengths weretested, all at a uniform elongationtime of 15 min. In this experiment,PABPN1 LALA behaved again neu-tral as expected with an RNA carry-ing anA15 tail (Fig. 6, lanes 1–8). Ontwo RNAs with A280 and A300 tails,mutant PABPN1 behaved as aninhibitor, just like on the RNA with

the A330 tail discussed above (Fig. 6, lanes 17-32). On an RNAwith an intermediate tail length of 110 nucleotides, PABPN1LALA by itself was unable to stimulate, as expected. When itwas used together with CPSF, the latter was still able to stimu-late polyadenylation, albeit slightly less efficiently than on itsown (Fig. 6, lanes 9–16). Thus, the inhibition of the CPSF-poly(A) polymerase interaction by PABPN1 occurs at a poly(A)tail length of more than 110 nucleotides and less than 280nucleotides. This supports the idea that this activity of PABPN1is at the heart of the length control mechanism.

FIGURE 5. PABPN1 inhibits poly(A) polymerase stimulation by CPSF on long tails. Polyadenylation reac-tions were set up containing, per unit volume of 20 �l, 80 fmol of RNA, 32 fmol poly(A) polymerase and, whereindicated, CPSF at a concentration approximately equimolar with RNA and/or 3.5 pmol of PABPN1 wild-type orLALA mutant. RNA, poly(A) polymerase, and PABPN1 were assembled on ice, and the mixture was prewarmedfor 10 min at 37 °C. Next CPSF was added, followed by a second preincubation for 15 min at 37 °C. 20-�l aliquotswere withdrawn for the 0-min time points, and then polyadenylation was started by the addition of ATP.Additional aliquots were withdrawn after 1, 3, and 5 min for L3pre-A15 (small light gray wedges, lanes 1–24) andafter 15, 30, and 60 min for L3pre-A330 (large dark gray wedges, lanes 25– 48). RNAs were resolved by polyacryl-amide gel electrophoresis. The sizes of DNA markers in nucleotides are indicated on the left.

Poly(A) Tail Length Control

AUGUST 21, 2009 • VOLUME 284 • NUMBER 34 JOURNAL OF BIOLOGICAL CHEMISTRY 22809

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

Stoichiometric Binding of PABPN1 IsRequired for Tail Length-dependentLoss of CPSFStimulation—The exper-iments described so far reveal thattermination of processive polyad-enylation is due to the inability ofCPSF to stimulate poly(A) polymer-ase when the two proteins are sepa-rated by a long poly(A) tail coveredby PABPN1. Loss of the interactionbetween poly(A) polymerase andCPSF is due to two properties ofPABPN1. First, on long tails, theprotein no longer supports theinteraction between poly(A) polym-erase at the 3� end and CPSF boundto the AAUAAA sequence, asinferred from the length-dependentinhibitory effect of the LALAmutant. Second, because PABPN1covers the poly(A) tail, it restrictsCPSF to the AAUAAA sequence,where it can no longer interact withpoly(A) polymerase. This interpre-tation predicts that the stimulationof poly(A) polymerase by CPSFshould be prevented only when thelong poly(A) tails are completelycovered by PABPN1. The predic-tionwas testedwith L3pre RNAcar-rying an A300 tail. Mutant PABPN1was titrated in a polyadenylationassay; approximately 3.6 pmol ofPABPN1 LALA were required formaximum inhibition of CPSF stim-ulation (Fig. 7A). In parallel, RNAbinding of the protein was assesseddirectly in a gel shift experiment,and a similar amount of PABPN1LALA was required to saturate theRNA (Fig. 7B). Thus, the disruptionof the CPSF-poly(A) polymeraseinteraction by PABPN1 requires thepoly(A) tail to be covered com-pletely. This also suggests that thelength of the poly(A) tail at whichthe termination of processive elon-gation occurs is counted byPABPN1.The Length Control Mechanism

Counts Only A Residues—Proces-sive elongation always generates atotal poly(A) tail length of � 250nucleotides, independently of thelength of the tail initially present.Thus, the length control mecha-nism is based on a determination ofthe total length of the poly(A) tail

FIGURE 6. Tail length dependence of the inhibitory effect of PABPN1. The reactions of 20 �l were assem-bled containing, as indicated, 80 fmol of L3pre-based RNAs with various poly(A) tail lengths, 32 fmol of PAP,CPSF (approximately equimolar with RNA), and increasing amounts of PABPN1 LALA. For each RNA a controlreaction containing a 10-fold amount of poly(A) polymerase was also carried out. After addition of RNA, poly(A)polymerase and PABPN1 the reaction mixtures were prewarmed for 10 min at 37 °C. After CPSF addition,prewarming was continued for 15 min, and then polyadenylation was started by ATP addition. The reactionswere stopped after 15 min at 37 °C, and the RNA was analyzed by gel electrophoresis. The amount of PABPN1LALA used corresponds to a calculated 1.3- and 2-fold saturation on L3preA280 and -A300, a 5- and 8-foldsaturation on L3preA110, and a 38- and 57-fold saturation on L3pre-A15 as judged by gel mobility shift analyses.

Poly(A) Tail Length Control

22810 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 34 • AUGUST 21, 2009

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

and not of the number of nucleo-tides added during the reaction orthe time poly(A) polymerase spendselongating the tail (45) (see also Fig.2).Which yardstick is used tomeas-ure tail length? PABPN1 is a likelycandidate, and this is supported bythe experiment shown in Fig. 7.Manipulation of the RNA substrateprovides an additional way to testthe hypothesis.Three different RNAs were com-

pared in processive polyadenylationreactions. One was the regular poly-adenylated L3pre RNA, L3preA15.The secondRNA, L3preN49A15, had49 nucleotides of additionalsequence of mixed compositioninserted before the beginning of thepoly(A) tail. The third RNA,L3preA15N63A15, had 63 nucleo-tides of mixed composition insertedbetween two stretches of 15 adeny-late residues each (Fig. 8A). Gel shiftexperiments showed that theseRNAs bound PABPN1 in propor-tion to the oligo(A) sequences inde-pendently of their positions (Ref. 41and data not shown). The non-Asequences inserted into the RNAsdo not bind PABPN1 (41). Uponincubation in the presence of CPSF,PABPN1, and poly(A) polymerase, arelatively small fraction of L3preA15RNA received poly(A) tails of up to�250 nucleotides within 30 s, inagreement with the processivenature of the reaction. Upon longerincubation, most of the remainingRNA was also polyadenylated, andmature tails continued to grow at alower rate, reflecting largely distrib-utive elongation beyond 250 nucle-otides (44, 45). After 5 min, the cen-ter of the length distributioncorresponded to �380 nucleotidesor slightly more than 300 nucleo-tides of poly(A) tail length (Fig. 8,A,middle panel, and B). Appropriatecontrols showed that the processivereaction depended on the simulta-neous stimulation of poly(A) poly-merase by both CPSF and PABPN1(Fig. 8A, middle panel). The sub-strate L3preN49A15 behaved verymuch like L3preA15, but at any timepoint products were �50 nucleo-tides longer (Fig. 8,A, left panel, and

FIGURE 7. PABPN1 has to coat long poly(A) tails to inhibit CPSF stimulation of poly(A) polymerase.A, titration of the inhibitory effect of PABPN1. 80 fmol of radioactive L3pre-A305 was incubated in a 20-�lreaction volume with 1.25 �g of competitor tRNA, 10 fmol of PAP, CPSF approximately equimolar to the RNA,and increasing amounts of PABPN1 LALA as indicated. After 5 min of preincubation at 37 °C, polyadenylationwas started by ATP addition and continued for 60 min at 37 °C. Some of the reactions were carried out twice(lanes 5 and 15 and lanes 4 and 16). A control reaction with a 10-fold amount of poly(A) polymerase was alsocarried out as explained under “Experimental Procedures.” RNA was analyzed by gel electrophoresis. The sizesof DNA markers are indicated on both sides. B, titration of PABPN1 LALA in a gel mobility shift assay. 80 fmol ofradioactive L3pre-A305, 1.5 �g of tRNA, and increasing amounts of PABPN1 LALA were incubated in 20-�lreaction volumes under polyadenylation conditions except that ATP was left out. After 30 min of incubation atroom temperature, complexes were resolved on a native polyacrylamide/agarose composite gel and detectedby phosphorimaging.

Poly(A) Tail Length Control

AUGUST 21, 2009 • VOLUME 284 • NUMBER 34 JOURNAL OF BIOLOGICAL CHEMISTRY 22811

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

B). Thus, the additional 49 nucleotides upstream of the poly(A)tail were not counted as part of the length control mechanism.In other words, the polyadenylation complex counts poly(A)length and not the distance between the 3� end and the CPSF-binding site. The substrate L3preA15N63A15 was also elongatedprocessively like L3preA15, but again at any time point theproducts were �50 nucleotides longer (Fig. 8, A,middle panel,and B). Although the limited precision of the length controlmechanism does not allow us to concludewhether theA15 tractseparated from the rest of the tail by 63 nucleotides of hetero-geneous sequence was counted as part of the tail or not, the N63sequence definitely was not counted. Together, the experi-ments show that only poly(A) is counted as part of the tail; othersequences are ignored even when they are surrounded bypoly(A). Because this matches the binding specificity ofPABPN1, the results suggest that binding of PABPN1, possiblyeven contiguous binding, is required for poly(A) counting andsupports the idea that a regular poly(A)-PABPN1 complexforms the basis of the counting mechanism.

DISCUSSION

Based on the data reported here and previously, the polyad-enylation reaction can be described as follows. In the normalsituation, when the RNA substrate is first cleaved and thenpolyadenylated, CPSF participates directly in the cleavage reac-tion, with the 73-kDa subunit acting as the endonuclease (51),and poly(A) polymerase is thought to be part of the cleavage

complex as well. In the uncoupled reaction investigated here,CPSF associates with the AAUAAA sequence of the precleavedRNA and recruits poly(A) polymerase, and poly(A) tail synthe-sis begins. The first molecule of PABPN1 binds to the RNAwhen the poly(A) tail has reached a length of 10–12 nucleotides(29). Additional copies bind as the tail gets longer. PABPN1binding has three effects. First, PABPN1preventsCPSFbindingto the poly(A) tail and restricts this protein to the AAUAAAsequence. This was shown both by elongation (Figs. 1 and 2)and by direct binding assays (Fig. 4). Second, PABPN1 facili-tates the interaction between CPSF bound to the AAUAAAsequence and poly(A) polymerase at the 3� end over the dis-tance of the poly(A) tail separating the two proteins, as shownby a constant rate of poly(A) tail elongation in the presence ofboth CPSF and PABPN1 (45) in contrast to a pronounced dis-tance dependence of elongation in the absence of PABPN1 (Fig.3). An interaction between CPSF and poly(A) polymerase canbe inferred from the CPSF dependence of polyadenylation andhas previously been shown by pull-down experiments (38). Pre-sumably, PABPN1 facilitates this interaction because the RNAhas to fold back on itself for the contact to bemaintained duringthe elongation reaction, and PABPN1 may promote this struc-ture (49). Third, a direct contact between PABPN1 and poly(A)polymerase provides additional stabilization of the polyadenyl-ation complex (41). The simultaneous interaction of poly(A)polymerase with both CPSF and PABPN1 endows the complex

FIGURE 8. Only poly(A) is counted as part of the tail: elongation of mixed tails. A, reaction mixtures were assembled on ice containing, per 25-�l unitvolume, 80 fmol of RNA, 1.25 �g of tRNA, 10 fmol of PAP, 1200 fmol of PABPN1, and an amount of CPSF approximately equimolar with RNA. Control reactionswith a 10-fold amount of PAP were done in parallel. The mixtures were preincubated for 2 min at 37 °C, and the reactions were started by the addition of ATP.Aliquots were withdrawn at the time points indicated. The reaction products were analyzed by gel electrophoresis. Unreacted substrate RNA is shown in thefirst lane of each of the three sets of reactions. The size of DNA markers (in nucleotides) are indicated on the right. B, line traces showing the RNA sizedistributions from polyadenylation reactions in A at the 5-min time points. The vertical gray line to the left of the 404-nt marker indicates the peak of the productlengths obtained by extension of L3pre-N49-A15 and L3pre-A15-N49-A15.

Poly(A) Tail Length Control

22812 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 34 • AUGUST 21, 2009

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

with sufficient stability that a complete poly(A) tail is synthe-sized in a single processive event. Once the poly(A) tail hasreached a length of �250 nucleotides, the structure of thepoly(A)-PABPN1 complex changes such that it disrupts or nolonger supports the interaction between CPSF and poly(A) po-lymerase (Fig. 5). Loss of this interaction terminates processivepolyadenylation. Subsequent elongation relies only on the stim-ulation by PABPN1 and is therefore poorly processive and rel-atively slow. Because the rate of poly(A) tail elongation beyond250 nucleotides is very similar in the presence of PABPN1 orCPSF or both, it was previously not possible to determinewhichof the two stimulatory interactions is lost. Only the mutationthat inactivated PABPN1 as a stimulator of poly(A) polymeraseuncovered the disruption of the CPSF-poly(A) polymeraseinteraction. The model of the polyadenylation reaction is sum-marized in the cartoon in Fig. 9.An alternative model dissociation or displacement of CPSF

upon completion of the poly(A) tail-was excluded by gel mobil-ity shift and UV cross-linking assays, which showed that CPSFremains bound to the RNA after termination of processiveelongation. A slight reduction in binding evident in bothtypes of assays is consistent with themodel because the bind-ing of CPSF to the RNA is stabilized by its interaction withpoly(A) polymerase, which no longer takes place under theseconditions.The length control mechanism measures the length of the

poly(A) tract and ignores other sequences, even if they areinserted between stretches of poly(A). Because PABPN1 is theonly protein in the reconstituted reaction that specifically rec-ognizes poly(A) and is used at concentrations sufficient to coverthe available poly(A) completely, it is hard to escape the con-clusion that this is the protein that counts the nucleotides of thetail. This model is also supported by the observation thatPABPN1 disrupts the poly(A) polymerase-CPSF contact onlywhen it covers the poly(A) tail completely. However, there is acaveat in the interpretation of this experiment, because com-plete coverage of the poly(A) tail by PABPN1 is also necessaryto prevent illegitimate binding of CPSF to the tail. Thus, theexperiment by itself does not permit an unambiguous conclu-sion about what is required to prevent the interaction betweenpoly(A) polymerase and CPSF when the latter is restricted tothe AAUAAA sequence.Additional support for a role of PABPN1 in measuring the

length of the poly(A) tail comes from the examination ofPABPN1-poly(A) complexes by electron microscopy, whichrevealed the existence of compact, spherical particles (49).These particles appeared self-limiting, never exceeding a diam-eter of 21 nm; on very long poly(A) molecules, several suchparticles of uniform size were arranged in a beads-on-a-stringpattern. Experiments with size-fractionated poly(A) showedthat one particle could accommodate 200–300 nucleotides ofpoly(A); shorter poly(A) supported the formation of smaller,FIGURE 9. Model of length control mechanism. CPSF binds the polyadenyl-

ation signal AAUAAA and recruits PAP. The first PABPN1 molecule joins thecomplex once the oligo(A) tail has reached a length of about 12 nucleotides.Additional PABPN1 molecules cover the growing tail. Formation of a tight,spherical PABPN1 particle on the growing poly(A) tail facilitates folding backof the RNA, which is required to maintain a contact between CPSF and poly(A)polymerase. Thus, the enzyme is held in the complex by cooperative interac-tions with both CPSF and PABPN1 and can synthesize the entire poly(A) tail ina processive manner. When the poly(A) tail exceeds a critical length of about

250 adenylate residues, additional PABPN1 molecules can no longer beaccommodated in the spherical RNA-protein complex, and the contactbetween poly(A) polymerase and CPSF cannot be maintained. Thus, duringfurther elongation of the poly(A) tail, poly(A) polymerase is held in the com-plex only by PABPN1; elongation becomes poorly processive and thereforeslow.

Poly(A) Tail Length Control

AUGUST 21, 2009 • VOLUME 284 • NUMBER 34 JOURNAL OF BIOLOGICAL CHEMISTRY 22813

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

presumably incomplete particles.We suggest that the sphericalparticles represent the folded back structure of the poly(A) tail,which is necessary to permit the interaction between CPSF andpoly(A) polymerase to bemaintained throughout chain growth.Poly(A) beyond a length of �250 nucleotides can no longer beaccommodated in the particle; thus poly(A) polymerase actingat the end of this RNA molecule cannot interact with CPSFanymore, and processive elongation stops (Fig. 9).Inherent size limits for linear structures like nucleic acids or

repetitive nucleic acid-protein complexes are hard to envision.A finite size is muchmore easily achieved with a globular struc-ture. Therefore, coiling up of the linear poly(A)-PABPN1 com-plex into a spherical complex seems like an attractivemodel forlength measurement. A much better understood case of lengthcontrol of a linear polymer by its accommodation in a globularprotein structure of defined size is the so-called “headful”mechanism of phage genome packaging into virions. One cap-sid can accommodate slightly more than one genome length ofDNA. Concatemeric phage DNA is spooled into an empty cap-sid until no more DNA can be accommodated, and then cleav-age of the genome is induced (52, 53). In contrast to the stableand well defined virion structures, the spherical poly(A)-PABPN1 complexes appear to be quite unstable (49). It will be achallenge to learn more about their structure.

Acknowledgment—We thank Gudrun Scholz for skillful technicalassistance.

REFERENCES1. Meyer, S., Temme, C., andWahle, E. (2004) Crit. Rev. Biochem. Mol. Biol.

39, 197–2162. Parker, R., and Song, H. (2004) Nat. Struct. Mol. Biol. 11, 121–1273. Wilson, T., and Treisman, R. (1988) Nature 336, 396–3994. Sachs, A. (2000) in Translational Control of Gene Expression (Sonenberg,

N., Hershey, J. W. B., and Mathews, M. B., eds) pp. 447–465, Cold SpringHarbor Laboratory Press, Cold Spring Harbor, New York

5. Tarun, S. Z., Jr., and Sachs, A. B. (1996) EMBO J. 15, 7168–71776. Hentze,M.W., Gebauer, F., and Preiss, T. (2007) inTranslational Control

in Biology and Medicine (Mathews, M. B., Sonenberg, N., and Hershey,J. W. B., eds) pp. 269–295, Cold Spring Harbor Laboratory Press, ColdSpring Harbor, New York

7. Amrani,N., Ghosh, S.,Mangus,D.A., and Jacobson,A. (2008)Nature453,1276–1280

8. Wickens, M., Goodwin, E. B., Kimble, J., Strickland, S., and Hentze, M.W.(2000) in Translational Control in Biology andMedicine (Mathews, M. B.,Sonenberg, N., and Hershey, J. W. B., eds) pp. 295–370, Cold Spring Har-bor Laboratory Press, Cold Spring Harbor, New York

9. Thompson, B., Wickens, M., and Kimble, J. (2007) in Translational Con-trol in Biology and Medicine (Mathews, M. B., Sonenberg, N., and Her-shey, J. W. B., eds) pp. 507–544, Cold Spring Harbor Laboratory Press,Cold Spring Harbor, New York

10. Klann, E., and Richter, J. D. (2007) in Translational Control in Biology andMedicine (Mathews, M. B., Sonenberg, N., and Hershey, J. W. B., eds) pp.485–506, Cold SpringHarbor Laboratory Press, Cold SpringHarbor, NewYork

11. Kwak, J. E., Drier, E., Barbee, S. A., Ramaswami, M., Yin, J. C., and Wick-ens, M. (2008) Proc. Natl. Acad. Sci. U.S.A. 105, 14644–14649

12. Saitoh, S., Chabes, A., McDonald, W. H., Thelander, L., Yates, J. R., andRussell, P. (2002) Cell 109, 563–573

13. Brown, C. E., and Sachs, A. B. (1998)Mol. Cell Biol. 18, 6548–6559

14. Groner, B., Hynes, N., and Phillips, S. (1974) Biochemistry 13, 5378–538315. Brawerman, G. (1981) CRC Crit. Rev. Biochem. 10, 1–3816. Rao, M. N., Chernokalskaya, E., and Schoenberg, D. R. (1996) Nucleic

Acids Res. 24, 4078–408317. Gu, H., Das Gupta, J., and Schoenberg, D. R. (1999) Proc. Natl. Acad. Sci.

U.S.A. 96, 8943–894818. Huarte, J., Stutz, A., O’Connell, M. L., Gubler, P., Belin, D., Darrow, A. L.,

Strickland, S., and Vassalli, J. D. (1992) Cell 69, 1021–103019. Colgan, D. F., and Manley, J. L. (1997) Genes Dev. 11, 2755–276620. Wahle, E., and Ruegsegger, U. (1999) FEMS Microbiol. Rev. 23, 277–29521. Zhao, J., Hyman, L., and Moore, C. (1999) Microbiol. Mol. Biol. Rev. 63,

405–44522. Shi, Y., Di Giammartino, D. C., Taylor, D., Sarkeshik, A., Rice,W. J., Yates,

J. R., 3rd, Frank, J., and Manley, J. L. (2009)Mol. Cell 33, 365–37623. Raabe, T., Bollum, F. J., and Manley, J. L. (1991) Nature 353, 229–23424. Wahle, E. (1991) J. Biol. Chem. 226, 3131–313925. Wahle, E., Martin, G., Schiltz, E., and Keller, W. (1991) EMBO J. 10,

4251–425726. Bienroth, S., Wahle, E., Suter-Crazzolara, C., and Keller, W. (1991) J. Biol.

Chem. 266, 19768–1977627. Keller, W., Bienroth, S., Lang, K. M., and Christofori, G. (1991) EMBO J.

10, 4241–424928. Nemeth, A., Krause, S., Blank, D., Jenny, A., Jeno, P., Lustig, A., andWahle,

E. (1995) Nucleic Acids Res. 23, 4034–404129. Wahle, E. (1991) Cell 66, 759–76830. Kuhn, U., and Wahle, E. (2004) Biochim. Biophys. Acta 1678, 67–8431. Keller, W., and Minvielle-Sebastia, L. (1997) Curr. Opinion Cell Biol. 9,

329–33632. Winstall, E., Sadowski, M., Kuhn, U., Wahle, E., and Sachs, A. B. (2000)

J. Biol. Chem. 275, 21817–2182633. Chen, Z., Li, Y., and Krug, R. M. (1999) EMBO J. 18, 2273–228334. Benoit, B., Mitou, G., Chartier, A., Temme, C., Zaessinger, S., Wahle, E.,

Busseau, I., and Simonelig, M. (2005) Dev. Cell 9, 511–52235. Jenny, A., and Keller, W. (1995) Nucleic Acids Res. 23, 2629–263536. Jenny, A.,Minvielle-Sebastia, L., Preker, P. J., and Keller,W. (1996) Science

274, 1514–151737. Murthy, K. G., and Manley, J. L. (1992) J. Biol. Chem. 267, 14804–1481138. Murthy, K. G., and Manley, J. L. (1995) Genes Dev. 9, 2672–268339. Jenny, A., Hauri, H. P., andKeller,W. (1994)Mol. Cell Biol. 14, 8183–819040. Kaufmann, I., Martin, G., Friedlein, A., Langen, H., and Keller, W. (2004)

EMBO J. 23, 616–62641. Kerwitz, Y., Kuhn, U., Lilie, H., Knoth, A., Scheuermann, T., Friedrich, H.,

Schwarz, E., and Wahle, E. (2003) EMBO J. 22, 3705–371442. Takagaki, Y., Ryner, L. C., and Manley, J. L. (1988) Cell 52, 731–74243. Christofori, G., and Keller, W. (1989)Mol. Cell Biol. 9, 193–20344. Bienroth, S., Keller, W., and Wahle, E. (1993) EMBO J. 12, 585–59445. Wahle, E. (1995) J. Biol. Chem. 270, 2800–280846. Sawicki, S. G., Jelinek, W., and Darnell, J. E. (1977) J. Mol. Biol. 113,

219–23547. Kuhn, U., Nemeth, A., Meyer, S., andWahle, E. (2003) J. Biol. Chem. 278,

16916–1692548. Martin, G., and Keller, W. (1996) EMBO J. 15, 2593–260349. Keller, R. W., Kuhn, U., Aragon, M., Bornikova, L., Wahle, E., and Bear,

D. G. (2000) J. Mol. Biol. 297, 569–58350. Wahle, E., Lustig, A., Jeno, P., and Maurer, P. (1993) J. Biol. Chem. 268,

2937–294551. Mandel, C. R., Kaneko, S., Zhang, H., Gebauer, D., Vethantham, V., Man-

ley, J. L., and Tong, L. (2006) Nature 444, 953–95652. Streisinger, G., Emrich, J., and Stahl., M. M. (1967) Proc. Natl. Acad. Sci.

U.S.A. 57, 292–29553. Lander, G. C., Tang, L., Casjens, S. R., Gilcrease, E. B., Prevelige, P., Polia-

kov, A., Potter, C. S., Carragher, B., and Johnson, J. E. (2006) Science 312,1791–1795

54. Matzura, O., and Wennborg, A. (1996) Comp. Appl. Biosci. 12, 247–249

Poly(A) Tail Length Control

22814 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 34 • AUGUST 21, 2009

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from

WahleUwe Kühn, Miriam Gündel, Anne Knoth, Yvonne Kerwitz, Sabine Rüdel and Elmar

Polyadenylation Specificity FactorRegulating the Interaction between Poly(A) Polymerase and the Cleavage and

Poly(A) Tail Length Is Controlled by the Nuclear Poly(A)-binding Protein

doi: 10.1074/jbc.M109.018226 originally published online June 9, 20092009, 284:22803-22814.J. Biol. Chem. 

  10.1074/jbc.M109.018226Access the most updated version of this article at doi:

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

Supplemental material:

  http://www.jbc.org/content/suppl/2009/06/09/M109.018226.DC1

  http://www.jbc.org/content/284/34/22803.full.html#ref-list-1

This article cites 49 references, 19 of which can be accessed free at

by guest on Novem

ber 9, 2020http://w

ww

.jbc.org/D

ownloaded from